zoom-in

Étoiles2

Forks0

Mis à jour8 avril 2026 à 10:03

refine ambiguous recognition by zooming into a selected region for closer analysis

Installation

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Exécuter dans Manus

Source

tenplusgood

tenplusgood/A-Harness

Ouvrir le dépôt GitHub Voir les dépôts du créateur

Téléchargement

Exécuter dans Manus

Métiers associésSOC

Basé sur la classification professionnelle SOC

Développeurs de logicielsProfessions informatiques et mathématiques·SOC 15-1252

SKILL.md

readonly

name	zoom-in
description	refine ambiguous recognition by zooming into a selected region for closer analysis
characteristics	["very fast local visual inspection tool","produces cropped reference image for reasoning only","helps resolve tiny/occluded target ambiguity"]
when_to_use	["target is small, far, blurry, or partially occluded","model needs closer inspection before choosing object part"]
when_not_to_use	["object and part are already clearly visible","do not pass zoomed image as detection input"]

Inputs:

image_path (string, required): absolute path to the RGB scene image.
object_name (string, required): The name of the object to be identified.

Behavior (for the model to follow, no scripts needed):

When the current recognition result is uncertain, you can select a local area, crop and zoom in on it, and reanalyze only that zoomed-in area.
Use the zoomed-in view to refine which object/part it is and update your reasoning before calling other skills (e.g. rex_omni, sam2).

Outputs (conceptual, no actual file I/O required):

refined_hypothesis (string): updated, more confident description of the object or part in the zoomed region.

Plus depuis ce dépôt

même dépôt

detection

tenplusgood/A-Harness

An integrated visual detection and segmentation tool that combines open-vocabulary object detection with prompt-driven precise segmentation capabilities, implementing a complete recognition pipeline from textual descriptions to pixel-level masks.

2026-04-082

dreamer

tenplusgood/A-Harness

An imagination-driven prompt writer that visualizes plausible physical interactions within an image to enhance contextual and functional reasoning for editing tasks.

2026-04-082

web-search

tenplusgood/A-Harness

Web search tool that retrieves concise textual knowledge for an affordance task: object parts, affordance mechanism, and how humans interact with the object. The decision model should CUSTOMIZE the search by providing a search strategy (search_focus, target_part, interaction_type) based on its analysis of the task. This produces much more targeted and useful results than a generic search. Returns: affordance_name, part_name, object_name, reasoning (text insights), plus searchable evidence (queries, results, crawled URLs).

2026-04-082

name	zoom-in
description	refine ambiguous recognition by zooming into a selected region for closer analysis
characteristics	["very fast local visual inspection tool","produces cropped reference image for reasoning only","helps resolve tiny/occluded target ambiguity"]
when_to_use	["target is small, far, blurry, or partially occluded","model needs closer inspection before choosing object part"]
when_not_to_use	["object and part are already clearly visible","do not pass zoomed image as detection input"]

Inputs:

image_path (string, required): absolute path to the RGB scene image.
object_name (string, required): The name of the object to be identified.

Behavior (for the model to follow, no scripts needed):

When the current recognition result is uncertain, you can select a local area, crop and zoom in on it, and reanalyze only that zoomed-in area.
Use the zoomed-in view to refine which object/part it is and update your reasoning before calling other skills (e.g. rex_omni, sam2).

Outputs (conceptual, no actual file I/O required):

refined_hypothesis (string): updated, more confident description of the object or part in the zoomed region.